A New Problem for Quantum Mechanics Alexander Meehan Abstract In this article I raise a new problem for quantum mechanics, which I call the control problem. Like the measurement problem, the control problem places a fundamental constraint on quantum theories. The characteristic feature of the problem is its focus on state preparation. In particular, whereas the measurement problem turns on a premise about the completeness of the quantum state (‘no hidden variables’), the control problem turns on a premise about our ability to prepare or control quantum states. After raising the problem, I discuss some applications. I suggest that it provides a useful new lens through which to view existing theories or interpretations, in part because it draws attention to aspects of those theories which the measurement problem does not (such as the role of conditional and relative states). I suggest that it also helps clarify the physical significance of the well-known no-go result—the no-cloning theorem—on which it is based. 1 Introduction 2 The Standard Measurement Problem 3 Quantum State Preparation 3.1 The preparation problem 4 The Control Problem 4.1 Set-up 4.2 Incompatibility argument 4.3 Other measurement problems Alexander Meehan A New Problem for Quantum Mechanics 5 Relation to Interpretations 5.1 Everettian mechanics 5.2 Bohmian mechanics 5.3 GRW theory 5.4 Other quantum theories 6 Relation to Foundations and the No-Cloning Theorem 6.1 Individual state determination 7 Conclusion Appendix 1 Introduction There is a compelling argument that it is impossible to reconcile Schrödinger’s quantum theory with the determinateness of measurement outcomes, unless we give up on the claim that the theory provides us with a complete description of physical systems. This is what I will call ‘the standard measurement problem’. More precisely, the problem is that the following assumptions, though individually innocent, are jointly untenable (Maudlin [1995]): (A1) The quantum state of a system determines, directly or indirectly, all of its physical properties [Completeness]. (A2) The quantum state of an isolated system always evolves in accord with a linear dynamical equation, such as the Schrödinger equation [Linearity]. (A3) Given determinate inputs, our measuring devices always produce unique, determinate outcomes [Determinate Outcome]. One desideratum for a quantum theory (or interpretation) is that it solve the standard measurement problem. A theory must reject at least one of the assumptions listed, and explain how the resulting gap is to be filled in—preferably, without introducing ‘measurement’ as a 2 Alexander Meehan A New Problem for Quantum Mechanics basic, unanalysed term in the dynamical laws.1 Bohmian mechanics, for instance, rejects (A1), GRW theory rejects (A2), and Everettian mechanics rejects (A3). I argue that there is another problem for quantum mechanics, distinct from the standard measurement problem. The characteristic feature of this new problem, which I call the control problem,2 is that it introduces considerations of state preparation on top of considerations of measurement. In particular, the problem is that the following claims are jointly incompatible: (B1) We can successfully prepare quantum states: at least some of our preparation devices are such that, if determinately fed many inputs, they output a non-trivial fraction of those inputs in some specified range of quantum states [Preparation] (here the ‘inputs’ are subsystems, and we define ‘the quantum state of a subsystem’ in the standard way, as its reduced state). (B2) The quantum state of an isolated system always evolves in accord with a deterministic dynamical equation that preserves the inner product, such as the Schrödinger equation [Unitarity]. (B3) It is always determinate whether or not a subsystem has been input into a given (measuring or preparation) device [Determinate Input]. (There is another background assumption needed to generate the incompatibility, which is also taken for granted in the standard measurement problem: (A4, B4) There is always, from a first-person view, an experienced ‘observed outcome’ of a measurement, even if—from a ‘God’s eye view’—there is no unique determinate outcome. Moreover, our measuring devices are at least somewhat informative and reliable. For example, if a spin-z measuring device is determinately fed an electron in the eigenstate "z, it will output ‘UP’ (and not ‘DOWN’) [Competent Measurement]. 1Authors often use the term ‘measurement problem’ to refer specifically to this deficiency of textbook collapse- on-measurement accounts. That is not how I am using the term here, but there is no real disagreement behind this difference; all agree that an account can reject one of (A1)–(A3), in this case (A2), and yet fail to be an acceptable fundamental physical theory for various reasons. For more on this terminological issue, see (Lewis [2007], pp. 750-1). For other presentations of the measurement problem that are similar to Maudlin’s above, see for example (Albert [1992]; Wallace [2012]; Healey [2017]; Myrvold [2018]). 2I thank an anonymous referee for suggesting this name. 3 Alexander Meehan A New Problem for Quantum Mechanics This assumption would be disastrous to deny, which is why it is background.) Another desideratum for a quantum theory is that it solve the control problem. A theory must reject at least one of the assumptions listed above and explain how the resulting gap is to be filled in—preferably, without introducing ‘measurement’ or ‘preparation’ as basic, unanalysed terms in the dynamical laws. GRW theory rejects (B2). Everettian mechanics rejects (B3). Bohmian mechanics, it turns out, rejects (B1) in favour of: (B1’) Given determinate inputs, our preparation devices can prepare those inputs in desired conditional quantum states (relativized to an extra parameter) but not desired quantum states simpliciter [Conditional Preparation]. Why are (B1)–(B4) incompatible? Briefly, the idea is to consider a certain type of preparation protocol, in which one chooses which quantum states to prepare based on which outcome of some initial measurement one (first-personally) observes. It is straightforward to show that if the assumptions were all true, then these protocols would violate a variant of the ‘no-cloning theorem’ (Wootters and Zurek [1982]) and so would be impossible to implement. But this cannot be right; we implement such protocols all the time. Therefore, at least one of the assumptions must be false. The main difference between the control problem and the standard measurement problem is the difference between (A1) and (B1). Whereas the standard measurement problem makes an assumption about what quantum states are (what physical properties they determine), the control problem makes an assumption about what we can do with quantum states—about the extent to which we can control the quantum states of the subsystems that we study in the laboratory. It is rare to find, for any interpretation or theory, detailed discussion of preparation and its limits. My hope is that the control problem will motivate such discussion. Is the control problem just a ‘non-standard’ version of the measurement problem? On the one hand, the measurement problem is often glossed as a problem with interpreting quantum mechanics as a complete physical theory, and most discussions involve some sort of state completeness assumption, suggesting Completeness is a defining feature of the problem. If that is the case, then solving the measurement problem does not entail solving the control problem; consider, for example, a theory which denies that quantum states specify 4 Alexander Meehan A New Problem for Quantum Mechanics macro-properties (like the readings on measuring devices), but maintains that we can prepare certain micro-systems in certain quantum states. On the other hand, the control problem invokes Competent Measurement, and crucially involves a measurement-type interaction (combined with a preparation). So if the measurement problem is defined as the general puzzle that measurement interactions can be problematic under Schrödinger dynamics, then the control problem should be viewed as a specific aspect of it. Of course, to some extent this question of how broadly to define ‘the measurement problem’ is a terminological one. What matters, for our purposes, is whether the control problem is of interest to the foundations of quantum mechanics. In this paper, I aim to show that it is. The plan is as follows. In Section 2, I review the standard measurement problem. In Section 3, I clarify Preparation and discuss whether it is already in tension with Unitarity. It turns out that this question has been raised under the header of the ‘preparation problem’ (Ruetsche [1996]; Wessels [1997]), so I take the opportunity to review this problem as well. In Section 4, I present the main result of the paper, the incompatibility argument for the control problem. I also compare the control problem with some non-standard versions of the measurement problem that have been raised (Maudlin [1995]; Gao [2019]). In Section 5, I turn to consider how different interpretations may respond to the control problem. The treatment here is necessarily brief, but some basic direction is given. Finally, in Section 6, I relate the problem to another foundational issue. There is a folklore in the literature that the no-cloning theorem rules out the possibility of ‘individual state determination’ (D’Ariano and Yuen [1996]; Vaidman [2015]). In this section, I argue that the folklore is mistaken, and suggest that the right result in the vicinity is actually a variant of the control problem. The payoff is a revised understanding of the physical significance of the no-cloning theorem. 2 The Standard Measurement Problem We begin by briefly recalling why (A1)–(A4) are jointly untenable. This will set the stage for my own incompatibility argument for (B1)–(B4), which has a similar set-up and structure. First, the set-up. Zarna is a physicist who specializes in quantum mechanics. Her laboratory 5 Alexander Meehan A New Problem for Quantum Mechanics is set up as shown in Figure 1.